Lauren Gardner (now at Johns Hopkins University) and I, along with Aleksa Zlojutro (UNSW Sydney) and Kamran Khan (University of Toronto) have just published this paper in The Lancet Infectious Diseases. It is a spatial risk analysis for measles in the United States. Our point of departure was the spatial analysis of the antivaccine movement in the U.S. published last year by Jacqueline Olive, Peter Hotez, and others published last year. That analysis considered vaccine refusal alone, correctly noting that it is becoming a global public health problem of critical significance.
However, for countries in which measles has been eliminated through vaccination campaigns (that is, it is no longer an endemic disease), including the U.S. since 2000, measles outbreaks occur because the virus is reintroduced by travelers from countries in which measles remains endemic. Thus the travel volume from such a country to any locality in the U.S. affects the risk of a measles outbreak there. So does the intensity of measles outbreaks in the country of origin of these travelers. Moreover, the size of the outbreak depends on the size of the susceptible population at that locality. A sophisticated risk analysis demands that all these risk factors be compounded.
Working at the county level (which is the most detailed resolution for which we could obtain data) we developed a risk function that incorporates all these risk factors. First, we computed the size of the susceptible population by multiplying the proportion of unvaccinated people with the county population size. The proportion of unvaccinated people can be estimated most precisely for counties in those seventeen states that still allow non-medical exemptions (NMEs) on a personal or philosophical basis. This the parameter used by Olive et al. (2018). It is slightly lower than the presumed proportion of unvaccinated people because it excludes those who claim religious exemptions and those that have legitimate medical exemptions. This is probably permissible because those numbers are typically much smaller. However, for many counties we do not have these data and, so, had to use less precise estimates of unvaccinated rates from surveys sponsored by the Centers for Disease Control (CDC).
Second, we multiplied the travel volume from a country with the size of any ongoing outbreak there for any year divided by its total population (that is, the incidence rate for measles in that country). The result is the expected number of infected people arriving from that country to a U.S. locality. Finally, we multiplied these two products (and, so, really we multiplied four factors: unvaccinated rate, destination U.S. county population, annual measles incidence rate for a foreign country, and travel volume from that country to that U.S. county). The we normalized this number by dividing all the final numbers by its greatest value. What we thus obtained is a risk measure that can be interpreted as the expected size of an oubreak in a given county. For instance, if this size is 100 for Cook County (with a value of 1.00 in our table), Los Angeles should expect an outbreak size of 43 (because it has the value 0.432 in our table).
Some counties have high risk largely because of the high rate of vaccine refusal in that region. These include: King County in Washington, Maricopa County in Arizona, Multnomah County in Oregon; and Travis County in Texas. Some counties have high risk because of both vaccine refusal and travel. These include: Harris and Tarrant Counties in Texas and Wayne County in Michigan. But there are also many counties, including our top four, that have high risk primarily because of a high volume of incoming travelers from foreign countries with high ongoing measles activity: Cook County in Illinois, Los Angeles County in California, Miami-Dade County in Florida, and Queens County in New York. Brooklyn, adjacent to Queens, has had by far the biggest outbreak so far. (We should emphasize that our calculation does not take public health responses into account which may well ward off large outbreaks in many high risk counties.) Here is the map for the situation in 2019:
We also calculated which countries present the greatest risk to the U.S. as a whole. These are: India, China, Mexico, Japan, Ukraine, Philippines, and Thailand. Of course, the limitations of our analysis should be pointed out. For instance, we miss Israel though it is the source of the Brooklyn outbreak.